Laws@SRI-AI.ARPA (04/09/84)
From: Ken Laws <Laws@SRI-AI.ARPA> The latest issue of IEEE Graphics (March '84) has an article comparing interpreted Prolog with compiled Pascal for a graphics application. The results are surprising to me. J.C. Gonzalez, M.H. Williams, and I.E. Aitchison of Heriot-Watt University report on the comparison in "Evaluation of the Effectiveness of Prolog for a CAD Application." They implemented a very simple set of graphical routines: much of the code is given in the article. They were building a 2-D entry and editing system where the polygons were stored as lists of vertices and edges. The user could enter new points and edit previously entered figures. This formed the front end to a system for constructing 3-D models from orthogonal 2-D orthographic projections (engineering drawings). Much of the code has the flavor of "For each line (or point or figure) satisfying given constraints, do the following ..." (Often only one entity would satisfy the constraints.) The authors report that the Prolog version (using assert and retract to manipulate the database) was more concise, more readable, and clearer than the Pascal version. The Prolog version also took less storage, was developed more quickly, and was developed with minimum error. What is more remarkable is that the interpreted Prolog ran about 25% faster than the compiled Pascal. They were using a PDP-11/34 with the NU7 Prolog interpreter from Edinburgh and the VU Pascal compiler from Vrije University.